Why /var/cache/git?

Why /var/cache/git?

[Bron Gondwana]

(resent - correct address for git list!)

(Gerrit: you're CC'd as the attribution in README.Debian
 which in the one place I can see /var/cache/git mentioned
 in the docs - without a justification for why it was
 chosen...)

I'm setting up shared repositories on a machine, and I notice
all the examples I can find on the web, and also the 
auto-created directory from the packages are /var/cache/git/

So I looked at the FHS, because that seemed odd to me:

http://www.pathname.com/fhs/pub/fhs-2.3.html#VARCACHEAPPLICATIONCACHEDATA

    /var/cache is intended for cached data from applications.
    Such data is locally generated as a result of time-consuming
    I/O or calculation. The application must be able to regenerate
    or restore the data. Unlike /var/spool, the cached files can
    be deleted without data loss.

I guess in theory, someone else has all those commits in their
local git repository somewhere.  Still, it feels to me that
/var/lib/git is the correct location.  Is there any reason why
/var/spool/git was chosen?

(I'm running Debian Lenny with the 1.7.1 backport packages, but
I see examples from Fedora with the same paths as well)

Bron.

Re: Why /var/cache/git?

[J.H.]

On 10/25/2010 03:30 AM, Bron Gondwana wrote:
> (resent - correct address for git list!)
> 
> (Gerrit: you're CC'd as the attribution in README.Debian
>  which in the one place I can see /var/cache/git mentioned
>  in the docs - without a justification for why it was
>  chosen...)

Couple of quick points:

$ grep -Ri "/var/git\|\/var\/git" *
Documentation/git-cvsserver.txt:     export
CVSROOT=:ext:user@server:/var/git/project.git

gitweb/README:    RewriteRule ^/var(/|/gitweb.cgi)?$	
/cgi-bin/gitweb.cgi [QSA,E=GITWEB_PROJECTROOT:/var/git/,L,PT]

gitweb/README:Second, repositories found in /pub/scm/ and /var/git/ will
be accesible

Are the only 3 instances where this is mentioned.  Looking at those
quick it was probably chosen based on the fact that it's a location that
likely exists on most machines.  Or just for explanatory reasons like
/example/dir/project.git could be used.

> I'm setting up shared repositories on a machine, and I notice
> all the examples I can find on the web, and also the 
> auto-created directory from the packages are /var/cache/git/

I can't speak for the web, it's a place filled with many many evil dirty
lies, but the primary documentation should likely get updated to not
actually shove perminant data into a transient location.  Bonus that
gitweb (caching) might use the directory for it's cache storage which
could cause some amusement.

> 
> So I looked at the FHS, because that seemed odd to me:
[...]
> I guess in theory, someone else has all those commits in their
> local git repository somewhere.  Still, it feels to me that
> /var/lib/git is the correct location.  Is there any reason why
> /var/spool/git was chosen?

Ummmm are you talking about for where a set of repositories exist on a
shared setup?  Personally, and depending on your beliefs as a sysadmin,
it should likely end up under something like /home/git or some other
shared location like /group/git, /pub/git, /shared/git or really
anywhere where you are likely to do backups and not just run rm -rf
because it should be a transient location.

> (I'm running Debian Lenny with the 1.7.1 backport packages, but
> I see examples from Fedora with the same paths as well)

Speaking from the Fedora angle (since I don't have a Debian box handy
anywhere or in a VM) the only places I'm seeing /var/git mentioned is:

/usr/share/doc/git-1.7.2.3/git-cvsserver.txt
/usr/share/doc/git-1.7.2.3/git-cvsserver.html
/usr/share/doc/git-cvs-1.7.2.3/git-cvsserver.txt
/usr/share/doc/git-cvs-1.7.2.3/git-cvsserver.html
/usr/share/doc/git-cvs-1.7.2.3/git-cvsserver.html
/usr/share/doc/gitweb-1.7.2.3/README
/usr/share/doc/gitweb-caching-1.6.5.2/README

which are the latest versions of all of those packages and it's obvious
it's the git-cvs and gitweb stuff that are to blame for the confusion there.

Probably easiest thing to do is adjust the documentation to be
/example/dir/project.git vs. /var/git since I agree, that's actually
dangerous.

- John 'Warthog9' Hawley

Re: Why /var/cache/git?

[Bron Gondwana]

On Mon, Oct 25, 2010 at 08:24:47AM -0700, J.H. wrote:
> On 10/25/2010 03:30 AM, Bron Gondwana wrote:
> > (resent - correct address for git list!)
> > 
> > (Gerrit: you're CC'd as the attribution in README.Debian
> >  which in the one place I can see /var/cache/git mentioned
> >  in the docs - without a justification for why it was
> >  chosen...)
> 
> Couple of quick points:
> 
> $ grep -Ri "/var/git\|\/var\/git" *
> Documentation/git-cvsserver.txt:     export
> CVSROOT=:ext:user@server:/var/git/project.git
> 
> gitweb/README:    RewriteRule ^/var(/|/gitweb.cgi)?$	
> /cgi-bin/gitweb.cgi [QSA,E=GITWEB_PROJECTROOT:/var/git/,L,PT]
> 
> gitweb/README:Second, repositories found in /pub/scm/ and /var/git/ will
> be accesible
> 
> Are the only 3 instances where this is mentioned.  Looking at those
> quick it was probably chosen based on the fact that it's a location that
> likely exists on most machines.  Or just for explanatory reasons like
> /example/dir/project.git could be used.

I have no problem with /var/git - it's /var/cache/git that worries
me, since the FHS quite clearly says that blowing away /var/cache/*
is not a problem.  It makes sense to back up /var/lib/ on any machine,
since that's somewhere that contains application data.  We already
back up a few things from /var/lib (like /var/lib/svn for example,
at least until we finish this switchover!)
 
Bron.

Re: Why /var/cache/git?

[Drew Northup]

On Tue, 2010-10-26 at 12:22 +1100, Bron Gondwana wrote:
> On Mon, Oct 25, 2010 at 08:24:47AM -0700, J.H. wrote:
> > On 10/25/2010 03:30 AM, Bron Gondwana wrote:
> > > (resent - correct address for git list!)
> > > 
> > > (Gerrit: you're CC'd as the attribution in README.Debian
> > >  which in the one place I can see /var/cache/git mentioned
> > >  in the docs - without a justification for why it was
> > >  chosen...)
> > 
> > Couple of quick points:
> > 
> > $ grep -Ri "/var/git\|\/var\/git" *
> > Documentation/git-cvsserver.txt:     export
> > CVSROOT=:ext:user@server:/var/git/project.git
> > 
> > gitweb/README:    RewriteRule ^/var(/|/gitweb.cgi)?$	
> > /cgi-bin/gitweb.cgi [QSA,E=GITWEB_PROJECTROOT:/var/git/,L,PT]
> > 
> > gitweb/README:Second, repositories found in /pub/scm/ and /var/git/ will
> > be accesible
> > 
> > Are the only 3 instances where this is mentioned.  Looking at those
> > quick it was probably chosen based on the fact that it's a location that
> > likely exists on most machines.  Or just for explanatory reasons like
> > /example/dir/project.git could be used.
> 
> I have no problem with /var/git - it's /var/cache/git that worries
> me, since the FHS quite clearly says that blowing away /var/cache/*
> is not a problem.  It makes sense to back up /var/lib/ on any machine,
> since that's somewhere that contains application data.  We already
> back up a few things from /var/lib (like /var/lib/svn for example,
> at least until we finish this switchover!)

Bron,
Doing a 'git grep "/var"' on the git sources and then searching that
output for "cache" turns up nothing. I think you'd best bring this up
with the Debian package maintainer. Debian is well known for thinking it
knows better than the original authors where stuff should be--and that's
why I don't run Debian as a server (too much time spent transliterating
paths just to make stuff work properly--when I'm lucky; it makes an
excellent desktop for beginners).

-- 
-Drew Northup N1XIM
   AKA RvnPhnx on OPN
________________________________________________
"As opposed to vegetable or mineral error?"
-John Pescatore, SANS NewsBites Vol. 12 Num. 59

Re: Why /var/cache/git?

[Gerrit Pape]

On Tue, Oct 26, 2010 at 09:21:39AM -0400, Drew Northup wrote:
> On Tue, 2010-10-26 at 12:22 +1100, Bron Gondwana wrote:
> > I have no problem with /var/git - it's /var/cache/git that worries
> > me, since the FHS quite clearly says that blowing away /var/cache/*
> > is not a problem.  It makes sense to back up /var/lib/ on any machine,
> > since that's somewhere that contains application data.  We already
> > back up a few things from /var/lib (like /var/lib/svn for example,
> > at least until we finish this switchover!)
> 
> Bron,
> Doing a 'git grep "/var"' on the git sources and then searching that
> output for "cache" turns up nothing. I think you'd best bring this up
> with the Debian package maintainer. Debian is well known for thinking it
> knows better than the original authors where stuff should be--and that's
> why I don't run Debian as a server (too much time spent transliterating
> paths just to make stuff work properly--when I'm lucky; it makes an
> excellent desktop for beginners).

Hi, it's also not my preference, but Debian adheres to the FHS.
Unfortunately /var/git is not allowed by the FHS, so a different
location must be chosen for Debian.  I chose /var/cache/git/, and
suggested to populate that directory with symlinks to the repositories
that should be available through git-daemon or gitweb, see
http://bugs.debian.org/483788

Regards, Gerrit.

Re: Why /var/cache/git?

[Jakub Narebski]

Gerrit Pape <pape@smarden.org> writes:
> On Tue, Oct 26, 2010 at 09:21:39AM -0400, Drew Northup wrote:
> > On Tue, 2010-10-26 at 12:22 +1100, Bron Gondwana wrote:

> > > I have no problem with /var/git - it's /var/cache/git that worries
> > > me, since the FHS quite clearly says that blowing away /var/cache/*
> > > is not a problem.  It makes sense to back up /var/lib/ on any machine,
> > > since that's somewhere that contains application data.  We already
> > > back up a few things from /var/lib (like /var/lib/svn for example,
> > > at least until we finish this switchover!)
> > 
> > Doing a 'git grep "/var"' on the git sources and then searching that
> > output for "cache" turns up nothing. I think you'd best bring this up
> > with the Debian package maintainer. Debian is well known for thinking it
> > knows better than the original authors where stuff should be--and that's
> > why I don't run Debian as a server (too much time spent transliterating
> > paths just to make stuff work properly--when I'm lucky; it makes an
> > excellent desktop for beginners).
> 
> Hi, it's also not my preference, but Debian adheres to the FHS.
> Unfortunately /var/git is not allowed by the FHS, so a different
> location must be chosen for Debian.  I chose /var/cache/git/, and
> suggested to populate that directory with symlinks to the repositories
> that should be available through git-daemon or gitweb, see
> http://bugs.debian.org/483788

Where did you get '/var/git' from?  "git grep /var/git" finds it
mentioned in one place in git-cvsserver manpage and in two places in
gitweb/README.  But git-daemon and gitweb's defaults are /pub/scm/git
or /pub/git - they are FHS-compatibile, aren't they?

So why Debian doesn't use either /pub/scm/git or /pub/git for 
git-daemon, git-cvsserver and gitweb?

-- 
Jakub Narebski
Poland
ShadeHawk on #git

Re: Why /var/cache/git?

[Junio C Hamano]

Gerrit Pape <pape@smarden.org> writes:

> Hi, it's also not my preference, but Debian adheres to the FHS.
> Unfortunately /var/git is not allowed by the FHS, so a different
> location must be chosen for Debian.

It seems that /var/lib/ seems to be more appropriate place to store
persistent database-y stuff. I see e.g. /var/lib/postgresql there.

Re: Why /var/cache/git?

[Gerrit Pape]

On Tue, Oct 26, 2010 at 11:30:04AM -0700, Junio C Hamano wrote:
> Gerrit Pape <pape@smarden.org> writes:
> > Hi, it's also not my preference, but Debian adheres to the FHS.
> > Unfortunately /var/git is not allowed by the FHS, so a different
> > location must be chosen for Debian.
> 
> It seems that /var/lib/ seems to be more appropriate place to store
> persistent database-y stuff. I see e.g. /var/lib/postgresql there.

I agree and re-opened
 http://bugs.debian.org/483788

Regards, Gerrit.

Re: Why /var/cache/git?

[Bron Gondwana]

On Tue, Oct 26, 2010 at 03:22:18PM +0000, Gerrit Pape wrote:
> On Tue, Oct 26, 2010 at 09:21:39AM -0400, Drew Northup wrote:
> > On Tue, 2010-10-26 at 12:22 +1100, Bron Gondwana wrote:
> > > I have no problem with /var/git - it's /var/cache/git that worries
> > > me, since the FHS quite clearly says that blowing away /var/cache/*
> > > is not a problem.  It makes sense to back up /var/lib/ on any machine,
> > > since that's somewhere that contains application data.  We already
> > > back up a few things from /var/lib (like /var/lib/svn for example,
> > > at least until we finish this switchover!)
> > 
> > Bron,
> > Doing a 'git grep "/var"' on the git sources and then searching that
> > output for "cache" turns up nothing. I think you'd best bring this up
> > with the Debian package maintainer. Debian is well known for thinking it
> > knows better than the original authors where stuff should be--and that's
> > why I don't run Debian as a server (too much time spent transliterating
> > paths just to make stuff work properly--when I'm lucky; it makes an
> > excellent desktop for beginners).
> 
> Hi, it's also not my preference, but Debian adheres to the FHS.
> Unfortunately /var/git is not allowed by the FHS, so a different
> location must be chosen for Debian.  I chose /var/cache/git/, and
> suggested to populate that directory with symlinks to the repositories
> that should be available through git-daemon or gitweb, see
> http://bugs.debian.org/483788

Maybe you'd better read the FHS again to double check that you're
adhering.

http://www.pathname.com/fhs/pub/fhs-2.3.html#VARCACHEAPPLICATIONCACHEDATA

Purpose:

  /var/cache is intended for cached data from applications. 
  Such data is locally generated as a result of time-consuming 
  I/O or calculation. The application must be able to regenerate 
  or restore the data. 

I fail to see how git can regenerate the symlinks...

Bron.